CFD Online Discussion Forums

CFD Online Discussion Forums (
-   FLUENT (
-   -   non-convergent computation - please help (

okorec June 13, 2009 06:28

non-convergent computation - please help

Im doing some work in Fluent 6.3. Im a student and this is my first time with CFD ever.

I did some reverse engineering on exhaust canal of cylinder head of combustion engine (3D scanning, surface reconstruction, model). I need to proof that model I got from that process is useful by having done a CFD simulation of air flow between valve and valve-seat and through the canal.

Well, the simulations are not doing good. Residuals oscillate and do not converge. I have no time for redoing the mesh (and I think that I wont be able to make a better mesh neither). Conditions: ideal-gas, RNG, K-epsilon, energy equation, discretization of second order, pressure on inlet/outlet 101325/96425 kPa, set turbulence and hydraulic parameter on inlet and outlet.

I tried to change some under relaxation factors and succeded in lowering the amplitude of mass-flow rate oscillation and a little of residuals oscillations. But since it is my first CFD case I am not always very sure what am I doing about these under relaxation factors. Afterall, oscillations are still present and the simulation do not converge.

Residuals of velocities oscillate about 0,003 and those of k and epsilon oscillate about 0,007. Reversed flow oscillates on outlet about 900 faces.

Model was meshed in Gambit. It has about 1,000,000 elements. In the critical place between valve and its seat is tgrid mesh of 1mm (or 0,5 mm for small valve openings) and the valve openings are from 2 to 9 mm.

If someone have a experience with something similar or just have a guess what to do, please help me. I have about a week to solve it :)

Thank you!

CTU in Prague

here I made a simple drawing of the situation. Actually it is a 3D.

-mAx- June 15, 2009 00:44

switch to double precision solver and re-try

okorec June 15, 2009 04:52

thank you max, but what do you mean by double precision solver?

my defined solver is:
pressure based
green-gauss node based
superficial velocity

or do you mean solution control?

I use second order do I need to use third order?


-mAx- June 15, 2009 06:44

when you start fluent, it will show you a window where you can choose between 2d, 2ddp, 3d and 3ddp
Choose the xddp (double precision)
If you start with command, then start with> fluent 3ddp

okorec June 16, 2009 04:40

well I tried it. but after 3500 iterations it does not seem to help :confused:

but thanks a lot for even trying to help me.


-mAx- June 16, 2009 05:12

If you stop to iterate, how is your pressure field and also your velocity field?
If they give non-realistic values, then try to initialize the pressure of the domain with 0.5*(pin-pout)
I also experienced some troubles with pressure inlet/pressure outlet

All times are GMT -4. The time now is 12:45.